Voxel51/TII-RATM-Drone-Racing

TII-RATM Drone Racing → FiftyOne (Native Multimodal MCAP)

The TII-RATM drone racing flights, as published with OpenVINS odometry in alvgaona/tii-ratm-rosbag2, converted to native multimodal MCAP episodes.

A quadrotor flies laps of a four-gate indoor track, three on an ellipse and three on a lemniscate. It carries a fisheye camera and a 500 Hz IMU, and a motion capture system watches the room throughout.

What makes the set useful is that both numbers are present: the estimate the drone computed from its own camera and IMU, and the reference the capture system measured at the same instant. Every episode carries both, plus the distance between them.

Installation

pip install fiftyone

Usage

import fiftyone as fo
import fiftyone.utils.huggingface as fouh

dataset = fouh.load_from_hub(
    "Voxel51/TII-RATM-Drone-Racing",
    name="TII-RATM-Drone-Racing",
    persistent=True,
)
fo.launch_app(dataset)

The flights the odometry found hardest:

view = dataset.sort_by("tracking_error_rmse_m", reverse=True)

What you get

Six flights, 10.9 minutes and 2,563 metres flown. Each episode carries:

• /camera, the onboard fisheye view at 640x480 and ~26 Hz, as foxglove.CompressedVideo
• /feature-tracks, the same view with the points the odometry is tracking drawn on it
• /camera-calibration, the equidistant fisheye intrinsics
• /imu.plot, accelerometer and gyroscope at ~490 Hz
• /pose-ground-truth, the motion capture pose
• /pose-vio, the odometry pose, in the same frame
• /tracking-error.plot, the distance between the two at each capture pose
• /trajectory, both flown paths as line strips
• /points-slam and /points-msckf, the feature points the estimator is holding
• /gates, the four track gates as boxes
• /gate-range.plot, the distance to each gate
• /tf, the odometry and camera frames
• /instruction, the track being flown

Across the whole set that comes to 17,305 camera frames, 9,658 feature-track frames, 327,086 IMU samples, 293,219 odometry poses and 179,283 capture poses.

Episodes carry the fields flight, track, distance_m, mean_speed_ms, max_speed_ms, tracking_error_rmse_m, tracking_error_median_m, tracking_error_max_m, tracking_error_final_m, num_camera_frames, num_feature_track_frames, num_imu_samples, num_vio_poses, num_ground_truth_poses, num_gates and duration.

Flight	Track	Distance	ATE RMSE	ATE max
flight-03p-ellipse	ellipse	380 m	0.612 m	2.349 m
flight-02p-ellipse	ellipse	446 m	0.634 m	3.455 m
flight-09p-lemniscate	lemniscate	447 m	0.642 m	1.699 m
flight-08p-lemniscate	lemniscate	367 m	0.674 m	2.194 m
flight-01p-ellipse	ellipse	479 m	1.072 m	3.482 m
flight-07p-lemniscate	lemniscate	444 m	1.656 m	3.784 m

Notes on the conversion

The odometry and the capture system report in unrelated frames, since the estimator starts at its own origin with an arbitrary yaw. A rigid transform is fitted over every time-matched pose in the flight and applied to the odometry, which is the usual way an absolute trajectory error is measured. Scale is not fitted: the estimate is metric because it is inertial-aided, so solving for scale would hide drift rather than measure it.

Because that fit spreads the residual across the whole flight, /tracking-error.plot does not start at zero and does not climb steadily. It is the distance from the reference at each instant, not the distance from a shared starting point. On flight-08p-lemniscate it runs 1.26 m at the start, dips to 0.15 m mid-flight and reaches 2.19 m by touchdown, where the estimate places the drone 1.4 m below the floor it has landed on.

Video is re-encoded to Annex-B H.264 without B-frames. The source carries raw bgr8 frames, which is most of its 44 GB.

The camera uses an equidistant (Kannala-Brandt) fisheye model. The intrinsics are published as they were calibrated, and a viewer that assumes a pinhole or plumb-bob model will not undistort this correctly.

Two source streams are not carried. loop_depth is typed as a 16-bit depth image but only a few dozen of its 307,200 pixels are ever set, so it is a sparse projection of loop-closure features rather than a depth map, and points_aruco is empty in every message of every flight.

License & attribution

The source release is distributed under CC BY 4.0, and this conversion is distributed under the same license.

Changes from the source: conversion to the FiftyOne MCAP flavor, re-encoding of the video to H.264, alignment of the odometry onto the capture frame, and encoding of the pose, IMU and detection streams as message streams.